Forensic Text Research Articles

A likelihood ratio-based evaluation of strength of authorship attribution evidence in SMS messages using N-grams

An experiment in forensic text comparison (FTC) within the likelihood ratio (LR) framework is described. The experiment attempts to determine the strength of authorship attribution evidence modelled with N-grams, which is perhaps one of the most basic automatic modelling techniques. The SMS messages of multiple authors selected from the SMS corpus compiled by the National University of Singapore were used for same- and different-author comparisons. I varied the number of words used for the N-gram modelling (200, 1000, 2000 or 3000 words), and then assessed the performance of each set. The performance of the LR-based FTC system was assessed with the log likelihood ratio cost (Cllr). It is shown in this study that N-grams can be employed within an LR framework to discriminate same-author and different-author SMS texts, but a fairly large amount of data are needed to do it well (i.e. to obtain Cllr < 0.75). It is concluded that the LR framework warrants further examination with different features and processing techniques.

Revelation and Rhetoric: A Critical Model of Forensic Discourse

Over the past thirty years or so, theoretical work in such fields as legal semiotics and law and literature has argued that the legal process is profoundly rhetorical. At the same time, a number of communication-based disciplines such as semiotics, sociolinguistics and linguistic anthropology have provided, particularly in interdisciplinary combination with law, a wealth of empirical evidence on, and insight into, the micro-contexts of language and communication in the legal process. However, while these invaluable nitty-gritty analyses provide empirical support for a rhetorical thesis, work in these areas has tended to ignore rhetoric as an explanatory principle. This article introduces an overarching rhetorical framework for the discursive construction and management of cases in contemporary Anglo-American legal processes. Taking ‘forensic’ as relating to the conduct of cases and ‘discourse’ as semiosis-in-practice, I argue that the practices within which forensic discourse is embedded are not, as the received legal view would have it, aimed at revealing an impartial truth but are deeply rhetorical practices aimed at persuading decision-makers to provide a remedy for a claimed wrong. By looking across forensic texts and contexts, I identify common elements of forensic discourse that can be found both in classical forensic orations and throughout the modern legal process and consider how these intersect with critical forces of agency and structure and the particularities of semiosis in situated context. An awareness of commonalities across forensic discourse can help sharpen our focus on the critical causes and consequences of individual and structural difference and point to consequential suggestions for reform.

On the Term “Dunamis” in Aristotle's Definition of Rhetoric

The term dunamis, by which Aristotle defines rhetoric in the first chapter of The Art of Rhetoric, is a “power” term, as its various meanings in Aristotle’s corpus—from vernacular ones like “political influence” to strictly philosophical ones like “potentiality”—attest.1 In the Rhetoric, however, dunamis is usually translated as “ability” or “faculty,” a designation that, compared to other terms that describe persuasion in ancient Greek poetics and rhetoric (such as “bia” [“force”] or “eros” [“seduction”]), marks rhetoric as a neutral human capacity rather than the use of language entangled in the vagaries of violence and desire.2 John Kirby calls Aristotle’s definition “one of the boldest moves in the history of the philosophy of language: to redefine rhetoric, not as the use of peitho but as the study of peitho” (1990, 227). The presumption of rhetoric’s ethical neutrality implied by dunamis has indeed become commonplace in interpretations of Aristotle’s treatise itself and of rhetoric as a social phenomenon. As George Kennedy puts it in his authoritative translation of the Rhetoric, “Aristotle was the first person to recognize clearly that rhetoric as an art of communication was morally neutral, that it could be used either for good or ill” (1991, ix). In this article, I would like to probe another, perhaps not so reassuring, implication of dunamis as a term for rhetoric—that as “an ability to see all available means of persuasion,” it does not need to become (or emulate) practical oratory. In what follows, I suggest that Aristotle’s terminology, however neutral it may appear, constitutes an intellectually and politically motivated act of naming that severs rhetorical knowledge from historically specific rhetorical practices and thereby erects a protective barrier between practical rationality and discourses of democracy.

Probabilistic Evaluation of SMS Messages as Forensic Evidence

This study is one of the first likelihood ratio-based forensic text comparison studies in forensic authorship analysis. The likelihood-ratio-based evaluation of scientific evidence has started being adopted in many disciplines of forensic evidence comparison sciences, such as DNA, handwriting, fingerprints, footwear, voice recording, etc., and it is largely accepted that this is the way to ensure the maximum accountability and transparency of the process. Due to its convenience and low cost, short message service (SMS) has been a very popular medium of communication for quite some time. Unfortunately, however, SMS messages are sometimes used for reprehensible purposes, e.g., communication between drug dealers and buyers, or in illicit acts such as extortion, fraud, scams, hoaxes, and false reports of terrorist threats. In this study, the author performs a likelihood-ratio-based forensic text comparison of SMS messages focusing on lexical features. The likelihood ratios (LRs) are calculated in Aitken and Lucy’s (2004) multivariate kernel density procedure, and are calibrated. The validity of the system is assessed based on the magnitude of the LRs using the log-likelihood-ratio cost (Cllr). The strength of the derived LRs is graphically presented in Tippett plots. The results of the current study are compared with those of previous studies.

Using ART1 Neural Networks for Clustering Computer Forensics Documents

Computer forensic text corpora are usually very heterogeneous and easily surpass the terabyte range. Classification methods should be an aid in the exploration of such corpora, but they do not help in the task of thematically grouping together documents. In this paper, we propose the use of Adaptive Resonance Theory (ART), applying the ART1 algorithm, to help in the task of thematically grouping together computer forensics documents into clusters. For the clustering approach we present the defined conceptual model and the software package implemented, in which a modified version of the ART1 algorithm was developed to improve the running time. Furthermore, real world forensic experiments were carried out to validate the model using a two-fold approach with a quantitative and a qualitative analysis method. The results demonstrate that our approach can generate good clusters when compared to the gold standard defined by domain area experts, with one clear advantage over other clustering methods (e.g. SOM and k-means) since there is no need to supply parameters beforehand such as the number of clusters. keywords computer forensics; document clustering; artificial neural networks; adaptive reasonance theory; ART1 algorithm Using ART1 Neural Networks for Clustering Computer Forensics Documents Georger Rommel Ferreira de Araujo(1) and Celia Ghedini Ralha(2) (1) Technical-Scientific Unit, Federal Police Department, Juazeiro do Norte, Brazil, rommel dot grfa at dpf dot gov dot br (2) Computer Science Department, University of Brasilia (UnB), Brasilia, Brazil, ghedini at cic dot unb dot br

The use of textual, grammatical and sociolinguistic evidence in forensic text comparison:

This article suggests the usefulness of the concept of ‘idiolectal style’ in forensic text comparison and presents evidence that was used by the defence in a forensic authorship attribution case to show that the author of four non-disputed faxes could also be the author of four disputed emails, both sets written in Spanish. Apart from qualitative textual analysis, two different quantitative approaches were undertaken: a) the use of the concepts of markedness and saliency in order to establish the rarity in the frequency and use of two grammatical variables - by comparing their behaviour to the one observed in a general corpus - indicated in both data sets the favourite use of two variants - redundant yo and compound periphrastic relative pronoun el cual, and b) the statistical analysis of sequences of linguistic categories classified and grouped the texts in the two sets under analysis (and also a control text set from another real forensic case) very closely, showing their discriminatory potential, in particular when the sequences were bigrams, something which would lead to the conclusion that these two text sets were produced by the same author.

Post-retrieval search hit clustering to improve information retrieval effectiveness: Two digital forensics case studies

This research extends text mining and information retrieval research to the digital forensic text string search process. Specifically, we used a self-organizing neural network (a Kohonen Self-Organizing Map) to conceptually cluster search hits retrieved during a real-world digital forensic investigation. We measured information retrieval effectiveness (e.g., precision, recall, and overhead) of the new approach and compared them against the current approach. The empirical results indicate that the clustering process significantly reduces information retrieval overhead of the digital forensic text string search process, which is currently a very burdensome endeavor.

Die vangnet van die woord: forensies-linguistiese getuienis in 'n lastersaak

The article takes the form of a report on a linguistic investigation of forensic texts which played a pivotal role in securing a conviction by the High Court of the Eastern Cape on a number of counts of criminal defamation in the case of the State versus Hoho. The report is introduced by a brief historical orientation of the field of forensic linguistics, both abroad (in particular in the USA), and its initial application in the South African context. A description is given of the run-up to, as well as the most salient details of, the case, informally known and reported in the media as the 'Father Punch case'. This is followed by a summary of the procedure according to which the linguistic data were processed, and then a presentation and discussion of some representative examples of the form in which the evidence was presented. Finally, an assessment is given of the potential value that the judgement could have on the future cooperation between the judiciary and linguists in South Africa.

Digital forensic text string searching: Improving information retrieval effectiveness by thematically clustering search results

Current digital forensic text string search tools use match and/or indexing algorithms to search digital evidence at the physical level to locate specific text strings. They are designed to achieve 100% query recall (i.e. find all instances of the text strings). Given the nature of the data set, this leads to an extremely high incidence of hits that are not relevant to investigative objectives. Although Internet search engines suffer similarly, they employ ranking algorithms to present the search results in a more effective and efficient manner from the user's perspective. Current digital forensic text string search tools fail to group and/or order search hits in a manner that appreciably improves the investigator's ability to get to the relevant hits first (or at least more quickly). This research proposes and empirically tests the feasibility and utility of post-retrieval clustering of digital forensic text string search results – specifically by using Kohonen Self-Organizing Maps, a self-organizing neural network approach. This paper is presented as a work-in-progress. A working tool has been developed and experimentation has begun. Findings regarding the feasibility and utility of the proposed approach will be presented at DFRWS 2007, as well as suggestions for follow-on research.

Corporate forensics class design with open source tools and live CDS

While Corporate and Law Enforcement (LE) Forensics share a common process; they differ in important ways including context, scope, outcomes, and tools. Since most forensic texts are created from a ...

On the use of corpora in the analysis of forensic texts

Twenty-five years ago Jan Svartvik published The Evans Statements: A Case For Forensic Linguistics in which he demonstrated that disputed parrs of a series of four statements, which had been dictated to police officers by Timothy Evans and incriminated him in the murder of his wife, had a measurably different grammatical style from the uncontested parrs and a new discipline was born. Initially its growth was slow; in unexpected places there appeared isolated articles in which the author, often a distinguished linguist, analysed disputed confessions, commented on purported records of interaction, evaluated the ability of ordinary people to understand legal language or challenged the authenticity of non-native-like language attributed to immigrants or aboriginals (see Levi, 1994 for references). There was, however, in those early days no attempt to establish a discipline or even a methodology for forensic linguistics - the work was usually undertaken as an intellectual challenge and almost always required the creation, rather than simply the application , of a method of analysis. By contrast, in the past five years, there has been a rapid growth in the frequency with which courts in a series of countries have called on linguists as expert witnesses, and, in consequence, there is now a developing methodology and a growing number of linguists who act as expert witnesses, a few even on a full-time basis (see Levi, this volume).

Forensic stylometry: A theoretical basis for further developments of practical methods

A theoretical examination shows that claims of near certainty of ascription by Morton's method, for determining authorship of disputed texts, are exaggerated. The findings of previous investigations into the method are briefly summarised. Specifically, its poor performance when applied experimentally to forensic texts is substantiated and then generalised by demonstrating that the observed lack of reliable discriminatory power is theoretically predictable. The main part of the paper consists of the derivation of a theoretical framework to provide a foundation from which to develop sound alternative approaches to overcome weaknesses inherent in Morton's method. The various components of one such practical proposal, which is of potential for forensic applications, are described.

Bullous Skin Lesions in Barbiturate Overdosage and Carbon Monoxide Poisoning

ABSTRACT To the Editor.— Mandy and Ackerman (213:253,1970) noted a lack of recognition in the United States of bullous skin lesions in persons with drug- or carbon-monoxide-induced coma. The authors overlooked the excellent description and photographs of these lesions in Legal Medicine: Pathology and Toxicology by Gonzales et al (New York, Appleton-Century-Crofts, 1954), and similar descriptions in other forensic pathology texts. Gonzales et al in this leading text attributed the lesions to hypoxia and pressure as have Doctors Mandy and Ackerman in their recent article.Unfortunately, in this country there is such great separation between the many physicians who study natural disease and death (at medical centers) and the few who study unnatural death (in medical examiners offices) that much knowledge obtained from investigation of death due to drugs, carbon monoxide, suicide, accidents, battered children, lead poisoning, etc, is not known to the medical community at large. This failing is